total_steps += config.summary_steps
print ("\r", "val [{}] loss: {:.3f}, Perplexity: {:.3f}, blue score: {:.3f} ".format(total_steps, val_loss, np.exp(val_loss), bleu_score))
# 儲存模型和結果
if total_steps % config.store_steps == 0 or total_steps >= config.num_steps:
save_model_t(model, optimizer, config.store_model_path, total_steps, timestr)
# save_model(model, optimizer, config.store_model_path, total_steps)
with open(f'{config.store_model_path}/output_{timestr}_{total_steps}.txt', 'w') as f:
for line in result:
print (line, file=f)
return train_losses, val_losses, bleu_scores
if __name__ == '__main__':
config = configurations()
print ('config:\n', vars(config))
train_losses, val_losses, bleu_scores = train_process(config)
timestr = time.strftime("%Y%m%d-%H-%M-%S")
plt.figure()
plt.plot(train_losses)
plt.xlabel('iters')
plt.ylabel('loss')
plt.title('train loss')
plt.savefig(f'./{config.figure_output_path}/train_loss_{timestr}.png')
plt.clf()
plt.figure()
resultsFileName = './e1/livello_{}/results_{}_{}_{}_keyboard_bis.txt'.format(level,af,opt,level)
finally:
line_storage.sort(key = lambda x: x.loss)
config.printResults(resultsFileName,line_storage,simulation_tStart,time(),af,opt,args["approximation"])
if args["p"]:
print("Plot best experiment loss history: \n{}".format(line_storage[0].toString()))
config.plot_history(line_storage[0].history)
return 0
if __name__ == '__main__':
import sys
import argparse
parser = argparse.ArgumentParser(description="Esperimento Livello 1")
parser.add_argument("activation-function", choices=['sigmoid','tanh','relu'],
help="funzioni di attivazione per la rete test")
parser.add_argument("optimizator", choices=['sgd', 'adagrad','adam'],
help="funzioni di ottimizzazione")
parser.add_argument("approximation", choices=["ideal","noref","ref"],
help="Approssimazione: ideale, senza riflessioni, con riflessioni",
default=0)
parser.add_argument("-p", action="store_true",
help="Alla fine della simulazione avvia la procedura di plot della loss")
args = vars(parser.parse_args())
#print(args)
config.configurations()
sys.exit(main(args))
from utils import *
from dataset import EN2CNDataset
from model import *
os.environ["CUDA_VISIBLE_DEVICES"] = '0'
seed = 1114
np.random.seed(seed)
random.seed(seed)
torch.cuda.empty_cache()
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
if __name__ == '__main__':
print("\nLoading configurations...", flush=True)
config = configurations(sys.argv)
config.load_model = True
print('config:\n', vars(config))
print("\nLoading data...", flush=True)
test_dataset = EN2CNDataset(config.data_path, config.seq_len, "testing")
test_loader = data.DataLoader(test_dataset, batch_size=1)
print("\nBuilding model...", flush=True)
model, _ = build_model(config, test_dataset.en_size, test_dataset.cn_size)
criterion = nn.CrossEntropyLoss(ignore_index=0)
print("\nStart testing...", flush=True)
test_loss, bleu_score, result = test_model(model, test_loader, criterion,
config.beam_width)